Four of the authors reply:
Nicolich and Gamble have questioned the methodology used in our recent study. 1 Our exposure assessment did not use contemporary estimates of NO2 and SO2 emissions, but used reconstructed emissions in three retrospective emission databases, representing emissions in the 1960s, 1970s, and 1980s, respectively. These estimated emissions were then input into atmospheric models to calculate a geographical distribution of pollutants. More details of the method used can be found in a companion paper. 2 In the next step, our goal was not to measure absolute concentrations in regions as small as 100 meters squared, as Nicolich and Gamble suggest. Instead, we measured the relative variation across a large region (Stockholm County) with a resolution for concentration gradients of 100 meters squared. Our data were reported to 100ths of a μg/m3 because the resulting estimated exposure was a continuous variable and cutpoints were determined by percentiles. To unequivocally state the cutpoints, as requested by the Journal Editors, it was necessary to use two decimals. Although two individuals on either side of a cutpoint are virtually indistinguishable, cutpoints at the group level are useful and quite valid for separating groups of exposed with different average exposures. In addition, we present analyses of continuous variables alongside the categorical analyses for a more complete picture. Similarly, while we appreciate the difficulty in measuring one drop in a train of tank cars, we were concerned chiefly with defining relative contrasts of exposure between individuals with regard to the pollutant in question, not with determining concentrations of the pollutant in relation to total air volume. In this regard, measurements to μg/m3 or parts thereof are perfectly feasible and are routinely used in pollution monitoring.
We agree with Nicolich and Gamble that there are important differences between our individual-level exposure estimates and workplace exposure studies. One does indeed have to assume that differences in concentration at the place of residence represent differences in total exposure reasonably well. For many pollutants, such as NO2 and fine particles, indoor concentrations reflect outdoor concentrations. Even for SO2, we believe the interindividual differences in estimated outdoor levels at the place of residence are a reasonable proxy for interindividual differences in exposure. Many studies have also shown that stationary monitoring can represent the ambient exposure of individuals reasonably well because it is less influenced by very local sources than is the case in workplace exposure situations. Furthermore, nondifferential misclassification of exposure due to imprecise exposure estimation would tend to attenuate any association between a continuous exposure variable and disease, and we find no reason to believe that exposure misclassification differed between cases and controls in our study. Therefore, the lack of association with SO2 may be questioned based on nondifferential misclassification, whereas for the association with NO2 the main potential problem due to nondifferential misclassification is that it may be underestimated.
The occupational analyses and choice of variables are based on careful analyses outlined in another companion paper. 3 Adjustment for occupation using three specific dichotomous exposure variables and one more general variable produces a finely tuned occupational adjustment that we believe is quite adequate. Whether the exposure “overwhelms” the ambient exposure is not the point; the strength of occupational exposures as potential confounders in the ambient analysis depends on both strong association with lung cancer risk and close correlation with the ambient exposures, a situation that is unlikely for any of the occupational exposures studied. In addition, exploratory analyses showed that inclusion or exclusion of occupational exposures in the model had little influence on the risk estimates for NO2.
1. Nyberg F, Gustavsson P, Järup L, Bellander T, Berglind N, Jakobsson R, Pershagen G. Urban air pollution and lung cancer in Stockholm. Epidemiology 2000; 11: 487–495.
2. Bellander T, Berglind N, Gustavsson P, et al. Using geographic information systems to assess individual historical exposure to air pollution in a population-based case-control study in Stockholm. Environ Health Perspect 2001; 109: 633–639.
3. Gustavsson P, Jakobsson R, Nyberg F, Pershagen G, Järup L, Schéele P. Occupational exposure and lung cancer risk—a population-based case-referent study in Sweden. Am J Epidemiol 2000; 152: 32–40.